JIT probe mechanism soon required for DFG and FTL OSR Exit

classic Classic list List threaded Threaded
5 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

JIT probe mechanism soon required for DFG and FTL OSR Exit

Mark Lam
FYI, I’m looking into making OSR exits work in a more memory efficient way using the JIT probe mechanism.  Once we make this transition, the DFG and FTL will not work for any CPU targets that don’t support the JIT probe mechanism.  The API to the JIT probe will probably change as well as this work progresses.  I’ll take care of updating all the probe implementations for ARM and x86 variants.

AFAICT, only CPU(MIPS) does not currently support the probe mechanism.  It’s up to the MIPS folks to implement that support if they want to continue to use the DFG.  Feel free to contact me if you have questions regarding the probe mechanism.

This work will be tracked in https://bugs.webkit.org/show_bug.cgi?id=174645.

Thanks.

Mark


_______________________________________________
webkit-dev mailing list
[hidden email]
https://lists.webkit.org/mailman/listinfo/webkit-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: JIT probe mechanism soon required for DFG and FTL OSR Exit

Konstantin Tokarev


19.07.2017, 02:49, "Mark Lam" <[hidden email]>:
> FYI, I’m looking into making OSR exits work in a more memory efficient way using the JIT probe mechanism.  Once we make this transition, the DFG and FTL will not work for any CPU targets that don’t support the JIT probe mechanism.  The API to the JIT probe will probably change as well as this work progresses.  I’ll take care of updating all the probe implementations for ARM and x86 variants.
>
> AFAICT, only CPU(MIPS) does not currently support the probe mechanism.  It’s up to the MIPS folks to implement that support if they want to continue to use the DFG.  Feel free to contact me if you have questions regarding the probe mechanism.
>
> This work will be tracked in https://bugs.webkit.org/show_bug.cgi?id=174645.

Could you elaborate, what code needs to be implemented for MIPS?

>
> Thanks.
>
> Mark
>
> ,
>
> _______________________________________________
> webkit-dev mailing list
> [hidden email]
> https://lists.webkit.org/mailman/listinfo/webkit-dev


-- 
Regards,
Konstantin
_______________________________________________
webkit-dev mailing list
[hidden email]
https://lists.webkit.org/mailman/listinfo/webkit-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: JIT probe mechanism soon required for DFG and FTL OSR Exit

Mark Lam


> On Jul 19, 2017, at 9:49 AM, Konstantin Tokarev <[hidden email]> wrote:
>
>
>
> 19.07.2017, 02:49, "Mark Lam" <[hidden email]>:
>> FYI, I’m looking into making OSR exits work in a more memory efficient way using the JIT probe mechanism.  Once we make this transition, the DFG and FTL will not work for any CPU targets that don’t support the JIT probe mechanism.  The API to the JIT probe will probably change as well as this work progresses.  I’ll take care of updating all the probe implementations for ARM and x86 variants.
>>
>> AFAICT, only CPU(MIPS) does not currently support the probe mechanism.  It’s up to the MIPS folks to implement that support if they want to continue to use the DFG.  Feel free to contact me if you have questions regarding the probe mechanism.
>>
>> This work will be tracked in https://bugs.webkit.org/show_bug.cgi?id=174645.
>
> Could you elaborate, what code needs to be implemented for MIPS?

See everything guarded by ENABLE(MASM_PROBE) in MacroAssemblerX86Common.cpp.  You’ll need to have the equivalent for MIPS.  Eventually, we may need some additional changes as well, but that depends on the solution currently being investigated.

Mark

>
>>
>> Thanks.
>>
>> Mark
>>
>> ,
>>
>> _______________________________________________
>> webkit-dev mailing list
>> [hidden email]
>> https://lists.webkit.org/mailman/listinfo/webkit-dev
>
>
> --
> Regards,
> Konstantin

_______________________________________________
webkit-dev mailing list
[hidden email]
https://lists.webkit.org/mailman/listinfo/webkit-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: JIT probe mechanism soon required for DFG and FTL OSR Exit

Mark Lam
FYI, following up on this work, I’ll soon land patches that relies on the probe mechanism being available.  So, I will now make the JIT probe (MASM_PROBE) a mandatory feature in order to enable the DFG and FTL (see https://bugs.webkit.org/show_bug.cgi?id=175446).  As a consequence, I’ll be disabling the DFG for the MIPS port in this patch.

Mark

On Jul 19, 2017, at 9:52 AM, Mark Lam <[hidden email]> wrote:



On Jul 19, 2017, at 9:49 AM, Konstantin Tokarev <[hidden email]> wrote:



19.07.2017, 02:49, "Mark Lam" <[hidden email]>:
FYI, I’m looking into making OSR exits work in a more memory efficient way using the JIT probe mechanism.  Once we make this transition, the DFG and FTL will not work for any CPU targets that don’t support the JIT probe mechanism.  The API to the JIT probe will probably change as well as this work progresses.  I’ll take care of updating all the probe implementations for ARM and x86 variants.

AFAICT, only CPU(MIPS) does not currently support the probe mechanism.  It’s up to the MIPS folks to implement that support if they want to continue to use the DFG.  Feel free to contact me if you have questions regarding the probe mechanism.

This work will be tracked in https://bugs.webkit.org/show_bug.cgi?id=174645.

Could you elaborate, what code needs to be implemented for MIPS?

See everything guarded by ENABLE(MASM_PROBE) in MacroAssemblerX86Common.cpp.  You’ll need to have the equivalent for MIPS.  Eventually, we may need some additional changes as well, but that depends on the solution currently being investigated.

Mark



Thanks.

Mark

,

_______________________________________________
webkit-dev mailing list
[hidden email]
https://lists.webkit.org/mailman/listinfo/webkit-dev


-- 
Regards,
Konstantin

_______________________________________________
webkit-dev mailing list
[hidden email]
https://lists.webkit.org/mailman/listinfo/webkit-dev


_______________________________________________
webkit-dev mailing list
[hidden email]
https://lists.webkit.org/mailman/listinfo/webkit-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: JIT probe mechanism soon required for DFG and FTL OSR Exit

Mark Lam
Looks like I have to disable DFG for the Windows port as well.  Relevant bugs:

Mark

On Aug 10, 2017, at 3:04 PM, Mark Lam <[hidden email]> wrote:

FYI, following up on this work, I’ll soon land patches that relies on the probe mechanism being available.  So, I will now make the JIT probe (MASM_PROBE) a mandatory feature in order to enable the DFG and FTL (see https://bugs.webkit.org/show_bug.cgi?id=175446).  As a consequence, I’ll be disabling the DFG for the MIPS port in this patch.

Mark

On Jul 19, 2017, at 9:52 AM, Mark Lam <[hidden email]> wrote:



On Jul 19, 2017, at 9:49 AM, Konstantin Tokarev <[hidden email]> wrote:



19.07.2017, 02:49, "Mark Lam" <[hidden email]>:
FYI, I’m looking into making OSR exits work in a more memory efficient way using the JIT probe mechanism.  Once we make this transition, the DFG and FTL will not work for any CPU targets that don’t support the JIT probe mechanism.  The API to the JIT probe will probably change as well as this work progresses.  I’ll take care of updating all the probe implementations for ARM and x86 variants.

AFAICT, only CPU(MIPS) does not currently support the probe mechanism.  It’s up to the MIPS folks to implement that support if they want to continue to use the DFG.  Feel free to contact me if you have questions regarding the probe mechanism.

This work will be tracked in https://bugs.webkit.org/show_bug.cgi?id=174645.

Could you elaborate, what code needs to be implemented for MIPS?

See everything guarded by ENABLE(MASM_PROBE) in MacroAssemblerX86Common.cpp.  You’ll need to have the equivalent for MIPS.  Eventually, we may need some additional changes as well, but that depends on the solution currently being investigated.

Mark



Thanks.

Mark

,

_______________________________________________
webkit-dev mailing list
[hidden email]
https://lists.webkit.org/mailman/listinfo/webkit-dev


-- 
Regards,
Konstantin

_______________________________________________
webkit-dev mailing list
[hidden email]
https://lists.webkit.org/mailman/listinfo/webkit-dev

_______________________________________________
webkit-dev mailing list
[hidden email]
https://lists.webkit.org/mailman/listinfo/webkit-dev


_______________________________________________
webkit-dev mailing list
[hidden email]
https://lists.webkit.org/mailman/listinfo/webkit-dev
Loading...